AshChadha-iitg/OpenMath

OpenMath is a personal or educational project (likely from an IIT Guwahati student) that demonstrates how to fine-tune a small model for math. With only 4 stars and 0 velocity, it lacks any market traction or community momentum. From a competitive standpoint, it possesses no unique moat; it uses a public dataset (NuminaMath) and a public base model (Qwen2) with standard fine-tuning libraries (PEFT/TRL). Frontier labs like OpenAI (o1-mini) and specialized open-source teams like DeepSeek (DeepSeek-Math) have already released models that far exceed the reasoning capabilities of a 0.5B parameter fine-tune. Furthermore, Alibaba's own Qwen2-Math series provides a much stronger baseline than this manual fine-tune. The high fork-to-star ratio (25 forks to 4 stars) often suggests a classroom or workshop setting where students fork a template, rather than organic developer adoption. The displacement horizon is '6 months' only because it is effectively already displaced by existing state-of-the-art open-source math models.